m_mineter@ed_ac_uk
Member
We are seeking to run a 20 member ensemble (B_20TR_CAM5_CN) with different starting points. Each case only require 72 cores, and would more readily be scheduled as say one 1440 core job, rather than 20 of 72 core jobs. I'm exploring how to restructure the ${CASE}.run and associated scripts to achieve this in an "ensemble.run" script, so one job can a) initiate and apirun each case in the ensemble; b) wait for all cases to run to completion; c) do postrun archiving etc; d) cause resubmission. Two thought arise: Has anyone already taken this approach? Has anyone already decided this is impractical?!Thanks!